Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. search_web and read_page have verb-noun naming (good), but descriptions lack WHEN/WHY guidance and parameter descriptions are minimal. No output schemas documented. Error handling is present but generic. Schema completeness is low, parameters lack type constraints, ranges, or format guidance. Descriptions are 100-150 chars (within baseline range) but lack actionable detail for LLM selection.

• Add explicit output schema to tool definitions: search_web returns {results: [{title: string, url: string, snippet: string}], total: number}. read_page returns {content: string, url: string, status: 'success'|'error'}.
• Expand parameter descriptions with format/constraint guidance: 'query (string, 1-500 chars): Search terms. Supports boolean operators (AND, OR, NOT). E.g., 'AI startups AND funding 2024'.' and 'url (string, http/https only): Full URL to fetch. Timeout 20s. Returns markdown-formatted content, max 12000 chars.'
• Document result limits and pagination: 'Returns up to 8 results per query. For more results, refine the query or use multiple searches with different keywords.'
• Add error classification and recovery guidance: 'If CAPTCHA is detected, wait 5 minutes and retry. If timeout occurs, try a simpler query or shorter URL. If 403 Forbidden, the site may block automated access, try a different source.'
• Clarify tool composition: 'Use search_web to discover URLs, then read_page to extract full content. search_web is fast but returns snippets; read_page is slower but returns complete markdown.'